[Piglit] Weekly 10 Picks from Patchwork for review and friendly reminder to clean out your old patches
maraeo at gmail.com
Thu Jun 25 04:50:31 PDT 2015
It would be nice if patchwork could filter patches according to touched files.
On Tue, Jun 23, 2015 at 11:53 AM, Daniel Vetter <daniel at ffwll.ch> wrote:
> On Fri, Jun 19, 2015 at 09:12:15PM +0100, Jose Fonseca wrote:
>> On 19/06/15 20:45, Ilia Mirkin wrote:
>> >On Fri, Jun 19, 2015 at 3:39 PM, Jose Fonseca <jfonseca at vmware.com> wrote:
>> >>On 19/06/15 13:32, Timothy Arceri wrote:
>> >>>Hi all,
>> >>>Unfortunately since its introduction patchwork hasn't seen a lot of love
>> >>>in the Piglit and Mesa projects so I thought I'd try something out to
>> >>>bring it out of the shadows and into the limelight.
>> >>>The idea is simple we have many useful but long forgotten patches
>> >>>sitting on the mailing list that would serve us much better sitting in
>> >>>the git repo, so once a week I (or anyone else that wants to help out)
>> >>>would pick 10 seemingly random older patches that could do with a
>> >>>I'm hoping this will help with both clearing out the backlog of patches
>> >>>and getting people thinking about patchwork.
>> >>>I'm interested in feedback on what people think about this idea.
>> >>Patchwork seems to fail to recognize submited patches. Eg. one of my
>> >>patches is https://patchwork.freedesktop.org/patch/51379/ but it has been
>> >>commited on
>> >>Why is that?
>> >It's very strict about matching patches. The diff has to be identical.
>> >Which it often isn't if you made minor changes, or rebased, or
>> Without a bit of fuzzy matching I'm afraid this looks a bit hopeless to me:
>> I believe the bulk of the patches are committed, and only a few is
>> forgotten. Looking at the patchwork backlog it's fair to say a large
>> portion of those committed don't get detected due to small changes. So the
>> end result is that developers have to click through and babysit the bulk of
>> their changes in patchwork, so that the few truly forgotten patches get to
>> stand out?
>> I don't think this will ever going to work. There's no incentive in the
>> system for the most prolific developers to spend so much of their time, for
>> the sake of the occasional contributor. The patchwork system seems bound to
>> echo what happens on the mailing list: their patches get to be lost twice...
>> There 's another concern -- one can only change the status of our own
>> patches. So if one commits on behalf of somebody else, and that patch
>> doesn't get recognized, one needs to get that other person to register and
>> click through patchwork?
>> I wonder if it wouldn't be better to have a more comprehensive solution for
>> review and tracking, ala github pull requests. Maybe have an official
>> mirror for mesa/piglit in github, or deploy gitlab
>> (https://about.gitlab.com/features/) in fdo.org, or something along those
>> lines, and start tracking this sort of things as pull requests.
>> I known it might look (and be) a wild idea at the moment, but I believe this
>> will be the future eventually: with things like cloud-based CI systems
>> (Travis CI, AppVeyor), projects can have testsuites run automatically on
>> pull requests (No GPU HW available, but one can still ensure builds don't
>> fail, run unit tests, and even rendering tests with SW renderers) and detect
>> issues even before reviewing or committing.
>> I've seen this happen first-handed: I once make a pull request to an
>> open-source project I had never contributed on github, a few minutes later
>> bot added a comment saying that the project built fine and all unit tests
>> passed, and all the maintainer had to do was clicking a button.
>> I'm now trying to repro this on some of my open source projects. (E.g,
>> Apitrace). I still have a long way to go, but already it is showing fruits
>> -- I immediately know when a Linux developr proposes a Apitrace change that
>> breaks Windows vuild (or a Windows developer breaks Linux build) , and I can
>> point them to the logs and they can often fix them selves. I hope one day I
>> have unit tests and more there too.
> We (the i915 kernel folks) are working on an improved patchwork (aiming to
> push it all to upstream patchwork ofc) which should address a lot of your
> concerns. It can trakc entire patch series, resends and has better
> automagic detection of when a related/rebased/slightly changed patch shows
> up or gets merged. Atm we're testing it internally a bit, but I hope we
> can roll it out to the public soonish. Damien is the one working on it
> Daniel Vetter
> Software Engineer, Intel Corporation
> Piglit mailing list
> Piglit at lists.freedesktop.org
More information about the Piglit